For the purposes of determining the amount of memory required for the computers running SQL Server 2005, first determine whether the planned deployment is small, medium, or large in terms of memory consumption.

Determine your deployment size by using the following table:

• If your deployment parameters are generally less than the listed values, your deployment can be considered small.

• If your deployment parameters are approximately equivalent to the listed values, your deployment can be considered medium.

• If your deployment parameters are generally greater than the upper limits of most of the listed values, your deployment can be considered large.

For SQL Server 2005, 4 gigabytes (GB) is the minimum required memory, 8 GB is recommended for medium size deployments, and 16 GB and greater above is recommended for large deployments.

Other factors that can influence your memory needs include:

• The use of SQL Server 2005 mirroring.

• The frequent use of files larger than 15 megabytes (MB).

On the server running SQL Server 2005, we recommend that the L2 cache per CPU have a minimum of 2 MB to improve memory.

Greater bus bandwidth helps improve reliability and performance. Consider that the disk is not the only user of bus bandwidth — for example, you must also account for network access.

The following list provides some best practices and recommendations for optimizing bus bandwidth.

• For medium to large-sized servers, greater bus bandwidth improves the system’s reliability, especially with added multi-pathing software. Conversely, greater bus bandwidth does not give a significant increase in reliability for smaller systems. The bus bandwidth’s reliability is improved through the redundant paths in the system and by avoiding single-point-of failure in hardware devices.

• Greater bus bandwidth provides improved performance in systems that frequently use large block transfers and sequential I/O.

• In smaller servers that use mostly sequential I/O, PCI becomes a bottleneck with three disks. For a small server that has eight disks performing mostly random I/O, PCI is sufficient. However, it is more common for PCI-X to be found on servers ranging from small to very large.

• Greater bus bandwidth is necessary to support a large number of disks.

• The capacity of bus bandwidth might be limited by the topology of the system. If the system uses direct attached disks, the number of slots limits the bus bandwidth capacity. However, for storage area network (SAN) systems, there is no physical limiting factor.

• More expensive servers typically have larger and faster buses. There is often no way to increase the capacity of the buses’ bandwidth without replacing the servers. However, the largest servers are more configurable. Consult with server providers for specifications.

The interfaces you use in your system can affect reliability and performance. Larger drives, all else being equal, increase mean seek time. Use the information in the following table to inform your choice of interface.

The disk topology that you use in your system can affect reliability and performance.

You should minimize latency on the I/O subsystem that serves the server running SQL Server 2005. Slow response from the I/O subsystem cannot be compensated for by adding other types of resources (for example, CPU or memory), but can influence and cause issues throughout the farm.

Use the information in the following table to inform your choice of topology.